Optimal ARMAX Model Order Identification of Dynamic Systems

This paper describes an effective approach for model order determination, which allows identifying the dynamical behavior of the mechanical system by using observation input-output data. The concept is based on the minimum means square error of the estimated transfer functions, which can effectively tackle measurement noise and modelling errorsto identify appropriate low-order transfer functions of the structures via an Auto-Regressive Moving Average eXogenous (ARMAX) model. The effectiveness of the proposed method is validated exclusively using experimental data obtained from a grinding test of an industrial manipulator SCOMPI robot. Some other common criteria,such as the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC), and the Noise Order Factor (NOF), are investigated in order to verify the performance of the propose  methodology. The results show that the proposed technique is cost-effective in terms of optimal model order determination, while the ARMAX model turns out to be the most appropriate representation for feature extraction at the low order. Thanks to its flexibility in handling model disturbance, the proposed optimization strategy is able to capture all the dominant oscillation modes of the structure at themlow orders, and system modal properties are efficiently and automatically determined, while the performance of the ARX model is shown to be less efficient when working at the low orders.

The authors declare no conflict of interest.

Not applicable

The datasets used in this study are openly available at [repository link] and the source code is available on GitHub at [GitHub link].

This work did not receive any external funding.